Genie 3
SE(3)-equivariant all-atom diffusion for scaffolds and binders
Genie 3 frames protein generation as reverse diffusion under an SE(3)-equivariant graph neural network that operates on all heavy atoms. Given length or conditioning constraints, it samples backbones and binders with high designability and structural plausibility. Its non-autoregressive sampling supports batched exploration of design space.
Genie 3 is being onboarded — request access and be first in line.
At a glance
- Input
- Length spec or target (PDB)
- Output
- Designed backbones (PDB)
- Developed by
- AlQuraishi Lab, Columbia University
- Published
- Jin et al., 2025 · 2025
#diffusion#se3-equivariant#backbone-generation
Related tools
More methods in Protein Design & Engineering
